论文部分内容阅读
飞行事故预测的目的在于预防事故。为提高事故预防的针对性和有效性,必须加强预测,以增强预防飞行事故的主动性。在ARIMA和SVM基础上,提出一种飞行事故组合预测方法。首先建立ARIMA模型,用以描述历史数据中的线性关系;然后,对ARIMA模型的残差构建SVM模型,用以模拟数据中的非线性规律,两者预测值之和就是最后的预测结果。美国空军1954—1993年飞行事故损坏飞机万时率的实证分析结果表明:利用该方法所建立的模型,能够对飞行事故作出较为准确的预测,模型精度总体优于单一的ARIMA或SVM模型。
The purpose of flight accident prediction is to prevent accidents. In order to improve the pertinence and effectiveness of accident prevention, it is necessary to strengthen the forecast so as to enhance the initiative of preventing flight accidents. Based on ARIMA and SVM, a combination forecasting method of flight accidents is proposed. Firstly, the ARIMA model is established to describe the linear relationship in the historical data. Then, the SVM model is constructed for the residuals of the ARIMA model to simulate the nonlinearity in the data. The sum of the two predicted values is the final prediction result. The empirical analysis of the damage rate of the aircraft handled by the U.S. Air Force from 1954 to 1993 shows that the model established by this method can make a more accurate prediction of flight accidents and the accuracy of the model is better than a single ARIMA or SVM model.